Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Show HN: AI-town, run your own custom AI world SIM with JavaScript

github.com

Post reply on HN
Print view View original

101–110 of 121 posts

Re: Show HN: AI-town, run your own custom AI world SIM with JavaScript

#101

Post by dalore »

drusepth wrote: post #43
pablo24602 wrote: post #31

I built something very similar when the paper came out- watch out for the instances where GPT returns both sides of the conversation (e.g. interpreting a prompt with chat history as being representative of what the output should look like, and putting out both "Jack" and "Jill"'s text). The first sample I went to suffered that problem (output from "Lucky's perspective below.) "Hey Stella! It's so great to see you! Yo…

Did you have any luck mitigating this problem?

You make the other side of the conversation the stopword. i.e. "Jill:" for Jack and Jill.

Re: Show HN: AI-town, run your own custom AI world SIM with JavaScript

#102

Post by empath-nirvana »

HeartStrings wrote: post #65

Could we not be living in a more advanced version of this same project?

There's a fairly direct analogy to be made between this project and Averroes' idea of the universal intellect, which has roots in Aristotle.

https://en.wikipedia.org/wiki/Averroes%27s_theory_of_the_uni...

> Averroes argues, as put by the historian of philosophy Peter Adamson, that "there is only one, single human capacity for human knowledge".[3] He calls it—using contemporary terminology—the "material intellect", which is one and the same for all human beings.[4] The intellect is eternal and continuously thinking about all that can be thought.[5] It uses faculties (e.g. the brain) of individual humans as a basis for its thinking process.[5] The process that happens in the human brain is called fikr by Averroes (known as cogitatio in Latin, often translated to "cogitation" in English), a process which contains not universal knowledge but "active consideration of particular things" that the person has encountered.[5] This use of human faculty explains why thinking can be an individual experience: if at one point the universal intellect is using one's brain to think about an object of thought, then that person is also experiencing the thinking.[5]

edit: I'm going to elaborate on this. In particular it's interesting to consider two kinds of knowledge -- the knowledge stored in ChatGPT's neural network, which is broad and "universal" and "timeless", and the knowledge of the individual agent's context, which is particular and local in both place and time. The "intellect" can't really do anything without being provided with a particular context in which to act.

Re: Show HN: AI-town, run your own custom AI world SIM with JavaScript

#103

Post by andai »

jmorgan wrote: post #3

If you haven't yet checked out the Generative Agents project referenced by OP, definitely give it a look, it's open source: https://github.com/joonspk-research/generative_agents Over the weekend Lance Martin got it working with local models using llama.cpp and ollama.ai which saves $ on longer sims since all inference happens locally https://twitter.com/RLanceMartin/status/1690829179615657985 . It's neat how the AI a…

I wonder how llama affects the simulation. Even llama2 consistently fails simple reasoning tasks.

(Though I suppose people aren't very reasonable, so it might give an accurate simulation!)

Re: Show HN: AI-town, run your own custom AI world SIM with JavaScript

#105

Post by pablo24602 »

dalore wrote: post #100
pablo24602 wrote: post #31

I built something very similar when the paper came out- watch out for the instances where GPT returns both sides of the conversation (e.g. interpreting a prompt with chat history as being representative of what the output should look like, and putting out both "Jack" and "Jill"'s text). The first sample I went to suffered that problem (output from "Lucky's perspective below.) "Hey Stella! It's so great to see you! Yo…

I did that but the trick to fix that was to use the other person's name and : prefix as the stop word. So in your case "Stella:". Then if it tries to continue the other side conversation it would just stop. I also noticed this when browsing the demo village, some conversations it ran through and did both parts.

This ended up being my solution as well- although I still ran into edge cases where GPT would output things like "Stella said". Never did manage to 100% fix this issue, even with creative prompts.

Re: Show HN: AI-town, run your own custom AI world SIM with JavaScript

#107

Post by cooper_ganglia »

lumenwrites wrote: post #79
icoder wrote: post #75

Earlier quoted context omitted.

Haha yes the most commonly used argument is (afaik) that if we could create a 'complete' virtual world, we would, and projects like this and how enthusiastic we are about it suggest that that's indeed the case. (The argument then of course continues that if we would, our world's inhabitants would some day too, ad infinitum. Given such a stack, the chance that we're actually still in the top layer is very small.)

My counterargument to this is that people simulating our world would be simulating all the bad stuff in it, which would be cruel and unethical. I think if humans were simulating us, they'd simulate a more utopian world. And if curious-but-indifferent aliens/robots simulated us, they wouldn't be wasting compute on simulating all the boring and uninteresting things we have in our world. I just don't see why would an in…

Assuming that our lived reality truly is just a simulation, I think it's important to also wonder why we would be being simulated. Not that we could ever hope to know or even comprehend our creator's intentions, but perhaps it's for research?

Could it be possible that they've just rented an RTX 40,090 and are simulating the entirety of our Universe to learn about how their (our?) species developed? Or how certain things would change if variables were altered? Perhaps there is a "multiverse" of infinite selves, but the multiverse is just a all the previous simulations that were tweaked slightly differently and had a different result.

Maybe changing the starting position of 1 atom in the beginning of the Universe has a butterfly effect that would entirely evade our assumptions. If we had the compute, why not simulate every possible reality? Perhaps there is 1 simulation that's occurred where everything was perfect, without anything cruel or unethical! And if we're gonna go as sci-fi as simulation theory, I don't think it's a far assumption that we could build things like fusion reactors or Dyson Spheres. We could learn so much, especially if the agents behaved naturally and had no idea that they themselves were simulated.

I don't actively believe we live in a simulation, but I think actively believing that we don't just doesn't make mathematical sense.

Re: Show HN: AI-town, run your own custom AI world SIM with JavaScript

#108

Post by all2 »

pyinstallwoes wrote: post #89
PUSH_AX wrote: post #74

Earlier quoted context omitted.

This just reminded me of Sam Harris’s arguments regarding free will not existing and certain thoughts that seemingly originate from nowhere. I guess an LLM is injecting them!

The Bible begins with a declaration of recursion stating it’s a LLM. In the beginning was the Word, and the Word was with God, and the Word was God.

That's John 1:1 in the New Testament.

The Bible begins in Genesis with a one week period of entropy reduction. "In the beginning God created the heavens and the earth", a summary of the following three or four chapters.

Re: Show HN: AI-town, run your own custom AI world SIM with JavaScript

#109

Post by jamesdwilson »

fnordpiglet wrote: post #70
jamesdwilson wrote: post #67

Earlier quoted context omitted.

Thanks! Can you please elaborate on "Cooperative agents have a lot of utility in more applied settings?"

Imagine being in your shop and telling a robot to get you a hammer. It can understand the instructions, look at the environment using an inverse image model, synthesize the environment with the instruction, delegate to path finders etc a route to the tool box and retrieve you a hammer without ever being programmed anything. Now abstract you to being other machine agents in the environment of a workshop being instruct…

Thank you very much @fnordpiglet for addressing my questions directly and thoroughly! Makes a lot of sense. I still don't understand why a16z wants us to do that though. what's in it for them?

Re: Show HN: AI-town, run your own custom AI world SIM with JavaScript

#110

Post by filchermcurr »

thih9 wrote: post #106

> a virtual town where AI characters live, chat and socialize The description says "live", "chat" and "socialize", but I only saw "chat". What exactly does "living" and "socializing" mean in this context?

I think by live it means exist and by socialize it means they walk around to other people.
Post reply on HN